Version 2.0

1 Executive Summary

Estimations relating to property counts, and consumption by unmetered customers, are used in the calculation of Leakage. The aim of this hypothesis is to identify potential inaccuracies in these estimations, which may affect the calculation of Leakage.

An average daily consumption for metered household and non-household customers has been derived. Consumption was then extrapolated to include non-metered customers, deriving a basic estimate of consumption for each DMA. This was then compared with Nightflow totals; it is expected that these would generally correlate, with higher Nightflow being recorded in DMAs with higher consumption. However, some DMAs were identified in which the Nightflow was significantly higher than would be expected based on estimated consumption, and these have therefore been highlighted for recommended further investigation; some of these appear to be related to discrepancies in different sources of property data.

In addition, Household Meter Readings have been analysed in order to determine if consumption patterns are different for customers using AMR meters, in comparison to other meter types. It was observed that AMR Meters generally record slightly higher consumption than other meter types, with older AMR meters recording less consumption than those less than five years old.

2 Experiments

2.1 Hypothesis 7.1: NHH Estimates

Currently, 46% of Non-Household (NHH) Customers are unmetered, and their water consumption is therefore estimated. This estimated affects the Leakage calculation. The purpose of this investigation is to use available data to derive a nominal consumption, and compare this to estimated Leakage data, in the context of property data discrepancies.

2.2 Hypothesis 7:2: ABP, Addresspoint and R&P Property Counts

The discrepancy between R&P properties and actual properties (derived from ABP and previously Addresspoint datasets) could have an impact in the leakage calculation; more specifically the night-use allowance.

R&P property counts are all the properties in the Southern Water billing system which is not geocoded. Addresspoint is a geocoded data set which has now been superseded by the Address Base Premium (ABP) dataset.

OFWAT Leakage Calculation

OFWAT Leakage Calculation

2.3 Hypothesis 7.4: AMR Meter Reporting

Approximately 10% of AMR Meters do not report readings. It is hypothesised that these may behave differently from other types of meter, and the estimated consumption of this 10% is leading to errors in the calculation of Leakage.

3 Data

The following datasets are used in the investigation of this hypothesis:

3.1 Consumption Data

This data was supplied by Southern Water, and covers the period from 1 April 2016 to 31 March 2019. The data is based on meter readings, with consumption derived based on the differences between readings.

At the start and ends of this three year period, the sample size becomes smaller, and therefore, investigations were focussed on the middle year of this data, 1 April 2017 to 31 March 2018, for which an average consumption per property, per day, has been derived. This is done based on a simple average of water used per day, spread evenly between meter readings. It does not account for seasonal variations in consumption at individiual property level, although given the large size of the dataset, and the fact that household meter readings are taken almost daily, this method is expected to give a general picture of seasonal consumption.

This data is divided into household and non-household customers. This data is then extrapolated to include unmetered customers, by taking the average customer consumption, per day, in each DMA, and applying this to each unmetered customer.

The results of this are displayed below. It should be noted that the scale of each graph is adjusted for clarity.

This data can be summarised more generally as follows.

It should be noted that this is a simple estimation of seasonal consumption, for the purposes of comparing this with nightflow data, and is not intended to replicate or replace the leakage calculation.

3.2 Household Meter Readings (2018-19)

This data is used to evaluate consumption by customers in comparison to customers with other types of meter. Consumption is compared accross different regions, as well as by the age of AMR meters, with meters over five years old classed as ‘Old’ and those more recently installed classed as ‘New’. This analysis includes only those where there exist multiple meter readings covering a time period of over 14 consecutive days.

The distribution of meter types is as follows:

Counts of Meter Types
New AMR Meters Old AMR Meters Other Meter Types
123061 418010 371956

3.3 Property Data

There is a 9% discrepancy between the R&P property count and ABP count. This has jumped from the 1% discrepancy from Addresspoint dataset. In order to reconcile the property counts, a DMA factor is applied to scale down the property count on a WBA level. There is a 8% difference between ABP and Addresspoint Factor applied.

Property Counts
WBA R&P Properties ABP Properties ABP DMA Factor Addresspoint Properties Addresspoint DMA Factor
Andover 32,778 35,675 0.92 32,193 1.02
Hampshire South 279,618 307,608 0.91 288,641 0.97
Hastings 54,068 58,946 0.92 50,386 1.07
Isle of Wight 72,124 83,195 0.87 74,332 0.97
Kingsclere 6,681 7,053 0.95 6,847 0.98
Medway 205,535 220,762 0.93 208,060 0.99
Sussex Coastal 253,587 285,080 0.89 249,925 1.01
Sussex North 116,044 125,763 0.92 117,898 0.98
Thanet 95,543 102,432 0.93 96,041 0.99
TOTAL 1,115,978 1,226,514 0.91 1,124,323 0.99

4 Results

4.1 Hypotheses 7.1 and 7.2: Estimated Consumption compared to Nightflow and Property Data

The estimated consumption, as derived in Section 3 above, was then plotted against the total Nightflow for the same DMA in the same period (2017-18). It would generally be expected that this comparison would show a high degree of correlation.

DMAs are coloured on this plot based on the discrepancy between R&P and ABP Property Data.

It can be seen from the above data that there are several significant outliers, in which the recorded nightflow is significantly higher than what would be expected, given the consumption estimates.

Therefore, the top 20 DMAs, ordered by total Nightflow, are shown below.

Property Counts
DMA Total Consumption (2017-18) Total Nightflow (2017-18) Property Discrepancy
SL30 310550.81 728460.1 1189
CR10 313703.96 676442.0 729
MS18 235282.90 607281.0 2918
OS11 117407.79 605713.0 3316
LS01 306169.10 498155.0 654
WG31 72328.57 472354.0 116
WM06 194102.43 459265.0 397
CW70 255752.86 448604.5 63
SL50 229640.17 448585.0 725
FL33 440197.02 423839.5 971
TW27 325288.69 422604.0 1036
MS23 353939.20 403868.0 1445
CW72 250145.90 380838.0 726
CR04 283824.72 377704.0 456
CR30 294027.99 348382.1 676
WL04 129368.09 347655.0 1186
DL10 174300.02 342350.0 472
SL31 237260.27 318062.5 98
BW16 248977.49 312230.0 781
AN10 242658.15 305574.0 373

Property Discrepancies (as per the scatter plot above) are identified on the map below. The top 20 DMAs, in terms of nightflow (as per the above table), are highlighted with a blue border. A number of these are known to be in an area of significant property development in North Kent.

4.2 Hypothesis 7.4: Consumption by Meter Type

Household Meter Readings data was divided into spatial areas (West, Central and East), with their consumption plotted over time.

With the exception of the Central region, it is notable that AMR Meters generally record a higher consumption than other meter types, with New AMR meters (those no more than five years of age) recording the highest consumption.

These results should be taken in the context of the WRC report, which summarised that:

  • Overall MUR (meter under registration) for 15mm household meters is 8.76%, showing a significant deterioration in performance compared to 2017.
  • Arad meters across Southern Water’s meter stock are estimated to be under-reading by an average 9.34%.
  • Other meters across Southern Water’s meter stock are estimated to be under-reading by an average of 7.18%.

If it is indeed the case that meters are under-recording, this likely to have a significant effect on the leakage calculation. As an example, it is assumed below that household consumption is, in fact, at the same level for all properties as it is for those with AMR meters less than five years old.

In this scenario, metered household consumption in the 2017-18 financial year would have been 2.23 Megalitres higher.

5 Recommendations

In some DMAs there is an apparent discrepancy between Consumption and Nightflow. Some of these appears to be as a result of property data discrepancies, most likely due to new property growth. However, some discrepancies betweeen Consumption, Nightflow and various property counts, may be for other reasons. It is therefore recommended that Southern Water investigate and comment on these. The findings of this would be used to direct future investigations into consumption estimates.

With regard to AMR meters, an enhancement could be made to the consumption estimates based on additional metering data, covering years prior to 2018. This may enable more accurate estimations of consumption. This could be combined with WRC Meter Under-Registration estimates, to estimate the effect of this on the Leakage calculation, in the manner as demonstrated in section 4.2.